The backport of #13150 to the 2.7 branch breaks the build on darwin. There is a special case for providing the _sysconfigdata module in _generate_posix_vars before get_platform is called (darwin only) and before the module is actually written.

However get_platform is called before, in site.addbuilddir(). Wondering if this call can either be avoided, or worked around.

options:

- backport issue586680 too (http://hg.python.org/cpython/rev/4742e7aea2f5/)
- fall back to generic values for osname, release, machine

any other options?

untested work-around. we don't need the builddir to generate the _sysconfigdata.py.

Backporting 4742e7aea2f5 would seem to be the simplest and most "correct" solution.

here's the backport, tested on Linux only. sorry, can't test this on MacOSX before Sunday.

Reverting the backport would also fix the breakage.

Is the addition of _sysconfigdata not a new feature? It changes the place where sysconfig (and distutils sysconfig?) looks for configuration data, and hence breaks all instructions that mention that you can change the Makefile in {sys.prefix}/lib/python2.7/config to globally affect the compilation flags (although there hopefully aren't a lot of those)

Guess what, I hadn't read the thread on this python-dev yet :-(

I backed it out until it can be done without breaking OSX. d174cb3f5b9e

Can landed corrected in 2.7.5.

tested the backport on MacOSX, fixing the build.

New changeset be3b4aa2ad28 by doko in branch '2.7':
- Issue #13150, #17512: sysconfig no longer parses the Makefile and config.h
http://hg.python.org/cpython/rev/be3b4aa2ad28